Volume 15: Low Carbon Cities and Urban Energy Systems: Part IV

Building Occupancy Prediction Through Machine Learning for Enhancing Energy Efficiency, Air Quality and Thermal Comfort: Review and Case Study Wuxia Zhang, Paige Wenbin Tien, John Calautit, Yupeng Wu



Recently, advanced prediction tools based on artificial intelligence are increasingly being employed for predicting occupancy patterns in buildings. The present work carries out a comprehensive review of studies using artificial intelligence and machine learning models to predict occupancy and its applications, covering studies about energy consumption, thermal comfort, lighting use and indoor air quality. The analyses show that while these studies have revealed that occupancy is a critical contributor to the energy prediction model, they have not paid enough attention to the thermal condition, air quality and their effect on occupant productivity and quality of life. In this study, occupancy detection with the vision-based camera is employed, which captures specific occupancy activities and other related behaviour like window opening behaviour. These activities will generate real-time deep learning influenced profile formation, which can train the prediction model. The results showed that the current CNN model framework provided an initial average detection accuracy of 84.48%. The ability of the deep learning detection to inform HVAC systems with significant help towards reducing building energy loads with the temperature setpoint changed. Therefore, it is important to develop an effective solution to increase the performance of buildings by assisting the HVAC control system in providing adequate indoor thermal comfort and air quality, while improving the building energy performance.

Keywords Machine learning, occupancy prediction, thermal comfort, energy consumption

Copyright ©
Energy Proceedings